Name: Memori MCP
Author: MemoriLabs

LABSCO SUMMARY

Recall before the reply, store after it — the loop is stated rather than left to whoever wires it up.

memori_recall belongs at the start of a turn and memori_advanced_augmentation after the response has been drafted, which puts the read before the reasoning and the write after it. That ordering is the whole integration: an agent calling neither has no memory, and one calling them in the wrong order stores context it has not used yet. What gets kept is described as durable facts and preferences rather than a transcript, so the store is meant to hold conclusions rather than history.

What it is

A memory server whose documentation names seven tools, built around one pair: retrieve the memories relevant to a query, and store durable facts and preferences from an exchange.

What you get

memori_recall retrieves the memories relevant to a query at the start of a user turn; memori_recall_summary fetches the broader memory state for session starts, daily briefs and project overviews.
memori_advanced_augmentation stores durable facts and preferences from the user's message and the assistant's response, meant to run after replying.
memori_compaction returns a structured brief after context compaction so an agent can resume; memori_feedback reports irrelevant, missing or stale memories; memori_quota checks usage and limits; memori_signup requests an account and API key when the user asks for one.

Requirements

A Memori API key from app.memorilabs.ai, set as MEMORI_API_KEY with an entity ID in MEMORI_ENTITY_ID; the hosted endpoint takes the same values as the X-Memori-API-Key and X-Memori-Entity-Id headers.

Setup effort

One command plus a key — npx -y @memorilabs/memori-mcp, then supply credentials

See installation options

USE CASES

01

Starting a turn with what you already know about someone

The recall runs against the query itself, so what comes back is the prior context relevant to this turn rather than everything on file.

02

Keeping what mattered from an exchange for next time

The write happens after the response is drafted, and what is stored is described as durable facts and preferences rather than the transcript.

7 tools

Listed in project docs. Your client may show a different list.

memori_recall

memori_recall_summary

memori_compaction

memori_advanced_augmentation

memori_feedback

memori_signup

memori_quota

BEFORE YOU INSTALL

What to know first

Tools you get

7 tools named in the project's documentation. The published npm package registers only recall and advanced_augmentation, without the memori_ prefix; the seven memori_ tools are available at https://api.memorilabs.ai/mcp/.

Requires

A Memori API key from app.memorilabs.ai (MEMORI_API_KEY)
An entity id for the end user (MEMORI_ENTITY_ID); a process id (MEMORI_PROCESS_ID) is optional
npx on PATH for the local package

INSTALLATION

Connect your client

Server setup

npx

TERMINAL

npx -y @memorilabs/memori-mcp

Hosted endpoint

ENDPOINT URL

https://api.memorilabs.ai/mcp/

Add this hosted endpoint to your MCP client using the credentials listed under Requirements.

MCP Clients

Cursor

CURSOR · JSON

{
  "mcpServers": {
    "memori": {
      "url": "https://api.memorilabs.ai/mcp/",
      "headers": {
        "X-Memori-API-Key": "${env:MEMORI_API_KEY}",
        "X-Memori-Entity-Id": "${env:MEMORI_ENTITY_ID}",
        "X-Memori-Process-Id": "${env:MEMORI_PROCESS_ID}"
      }
    }
  }
}

Save as ~/.cursor/mcp.json (or .cursor/mcp.json in the project) and restart Cursor. Export MEMORI_API_KEY (from app.memorilabs.ai), MEMORI_ENTITY_ID (a stable id for the end user, e.g. user_123) and optionally MEMORI_PROCESS_ID before starting the client, or put the values straight into the headers.

Claude Code

CLAUDE CODE · TERMINAL

claude mcp add --transport http memori https://api.memorilabs.ai/mcp/ \
  --header "X-Memori-API-Key: ${MEMORI_API_KEY}" \
  --header "X-Memori-Entity-Id: ${MEMORI_ENTITY_ID}" \
  --header "X-Memori-Process-Id: ${MEMORI_PROCESS_ID}"

Export MEMORI_API_KEY (from app.memorilabs.ai), MEMORI_ENTITY_ID (a stable id for the end user, e.g. user_123) and optionally MEMORI_PROCESS_ID before starting the client, or put the values straight into the headers. Run /mcp inside Claude Code to check the server status.

Codex

CODEX · TOML

[mcp_servers.memori]
enabled = true
url = "https://api.memorilabs.ai/mcp/"

[mcp_servers.memori.http_headers]
X-Memori-API-Key = "${MEMORI_API_KEY}"
X-Memori-Entity-Id = "${MEMORI_ENTITY_ID}"
X-Memori-Process-Id = "${MEMORI_PROCESS_ID}"

Add to ~/.codex/config.toml. X-Memori-Process-Id is optional.

Antigravity

ANTIGRAVITY · JSON

{
  "mcpServers": {
    "memori": {
      "serverUrl": "https://api.memorilabs.ai/mcp/",
      "headers": {
        "X-Memori-API-Key": "<YOUR_MEMORI_API_KEY>",
        "X-Memori-Entity-Id": "user_123",
        "X-Memori-Process-Id": "my_agent"
      }
    }
  }
}

In Manage MCP Servers, edit mcp_config.json, then restart Antigravity. user_123 and my_agent are samples: the end user's id and a name for the agent (optional).

Any MCP client — npx

ANY MCP CLIENT — NPX · JSON

{
  "mcpServers": {
    "memori": {
      "command": "npx",
      "args": [
        "-y",
        "@memorilabs/memori-mcp"
      ],
      "env": {
        "MEMORI_API_KEY": "<YOUR_MEMORI_API_KEY>",
        "MEMORI_ENTITY_ID": "user_123",
        "MEMORI_PROCESS_ID": "my_agent"
      }
    }
  }
}

Local stdio package. The values are your Memori API key, the end user's id and (optionally) an agent name. The npm package registers only recall and advanced_augmentation; the full seven tools are on the hosted endpoint.

Memori MCP